List alert delivery records across every check in the project

Returns delivery records project-wide, newest-id first, keyset-paged — the console's Delivery log tab and any automation that wants to confirm whether a given monitor's alert reached a destination.

A deliveries row is one (incident event, channel) outcome: pending while a delivery attempt is in flight, delivered on success, dead once the per-channel attempt ceiling is reached, or suppressed when it was deliberately not sent: the channel's rate cap dropped it, or it is a blocked alert whose run was answered within the hold (reason in last_error). incident_id correlates the row to GET /api/v1/incidents/{id} by matching (check_id, incident_opened_at) — it is null once the parent incident has aged out of the 1-year incidents retention window (see /privacy), or for an event type with no incident to correlate with (e.g. every-run). channel_name/channel_kind are null when the destination channel has since been deleted; the delivery record itself is not affected by a channel's deletion.

Pagination is keyset, on id descending. next_cursor is an opaque token; pass it back as cursor verbatim to fetch the next page, and stop when next_cursor is null. A cursor this API did not issue (or one that has been altered) is rejected with 400 INVALID_CURSOR.

?since= bounds incident_opened_at and defaults to the last 30 days; it is clamped to 365 days back regardless of what is asked for, because deliveries are pruned at 365 days (paired with the incidents retention window — see /privacy) and an earlier bound finds nothing further anyway.

Console stability note. This response is shaped by a console screen: the fields are the ones that screen draws, in the grouping it draws them in. It stays published because reading it is a reasonable thing to automate, and no field will be removed or change meaning without a version. Fields may be ADDED as the screen grows, so treat unknown keys as normal rather than as an error.